Image and video segmentation is one of the most fundamental yet challenging problems in computer vision. Dividing an image into meaningful regions requires a high level interpretation of the image that cannot be satisfactorily solved by only looking for homogeneous areas in an image. In the era of big data and vast computing power, one approach to model high level interpretation of images has been to use powerful machine-learning tools on huge annotated databases. While significant advances have been made in recent years, automatic image segmentation is still far from providing accurate results in a generic scenario. The creator of a video may desire to add information or a link to an object in a video, and may wish the added information or link to remain associated with that object throughout a video sequence.